Novel Image Matching Algorithm Based on Symmetrical Color-Spatial Feature
Bing Zhou · Mini-micro Systems · 2005
Color is the most common used feature for image matching in CBIR (Content Based Image Retrieval ) system by now. It is ve ry necessary to provide spatial information of color for increasing the accuracy of image matching based on color feature. The advantage of color feature is ind ependent of image distortion such as resizing and rotating. However after combin ing with spatial information, this advantage would be disappear. This paper prop osed a novel image-matching algorithm that based on symmetrical color-sp atial feature. By using the symmetrical spatial information for color feature, t his algorithm reserved the independence of color feature and eliminates the infl uence of distortion to the image matching while increasing the accuracy. Exp eriments show that the new algorithm reduces the mismatch caused by image distor tion especially the symmetrical distortion.